DOCSLIB.ORG

A Framework for Governmental Use of Machine Learning

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
A Framework for Governmental Use of Machine Learning REPORT FOR THE ADMINISTRATIVE CONFERENCE OF THE UNITED STATES A FRAMEWORK FOR GOVERNMENTAL USE OF MACHINE LEARNING Cary Coglianese University of Pennsylvania Law School This report was prepared for the consideration of the Administrative Conference of the United States. It does not necessarily reflect the views of the Conference (including its Council, committees, or members). Recommended Citation Cary Coglianese, A Framework for Governmental Use of Machine Learning (Dec. 8, 2020) (report to the Admin. Conf. of the U.S.), https://www.acus.gov/sites/default/files/documents/ Coglianese%20ACUS%20Final%20Report.pdf A Framework for Governmental Use of Machine Learning Cary Coglianese* TABLE OF CONTENTS INTRODUCTION ........................................................................................................ 3 I. LIMITATIONS OF HUMAN DECISION-MAKING ....................................................... 8 A. Physical limitations ...................................................................................... 10 1. Memory Capacity. ..................................................................................... 10 2. Fatigue. ..................................................................................................... 11 3. Aging. ........................................................................................................ 12 4. Impulse Control. ....................................................................................... 13 5. Perceptual Inaccuracies. .......................................................................... 14 B. Biases ........................................................................................................... 15 1. Endowment Effect. .................................................................................... 15 2. Loss Aversion. ........................................................................................... 16 3. System Neglect. ......................................................................................... 16 4. Hindsight Bias. .......................................................................................... 16 5. Availability Bias. ....................................................................................... 17 6. Confirmation Bias. .................................................................................... 18 7. Framing..................................................................................................... 19 8. Anchoring. ................................................................................................. 19 9. Susceptibility to Overpersuasion. ............................................................. 19 10. Implicit Racial and Gender Biases. ........................................................ 20 C. Problems with Group Decision-Making ...................................................... 21 D. Implications for Decision-Making in Government ...................................... 22 * Edward B. Shils Professor of Law and Director of the Penn Program on Regulation, University of Pennsylvania Law School. This report was prepared for the Administrative Conference of the United States (ACUS) Office of the Chairman project on Agency Decision Making by Artificial Intelligence. The author gratefully acknowledges major contributions to Parts I and II by Alicia Lai, as well as many constructive contributions both to this project and related collaborations by Steven Appel, Lavi Ben Dor, and David Lehr. Helpful comments from Richard Berk, David Rubenstein, and members of the ACUS Committee on Agency Use of Artificial Intelligence are appreciated. Emma Ronzetti and Roshie Xing provided valuable research assistance.The views expressed here are those of the author only and not necessarily of ACUS or its staff or members. 1 II. MACHINE LEARNING’S PROMISE FOR IMPROVING GOVERNMENTAL DECISION- MAKING ................................................................................................................. 23 A. What Makes Machine Learning Different? ................................................. 25 B. Machine Learning’s Advantages in Private-Sector and Medical Decision- Making .............................................................................................................. 27 1. Accuracy. .................................................................................................. 27 2. Capacity. ................................................................................................... 29 3. Speed. ........................................................................................................ 30 4. Consistency. .............................................................................................. 30 C. Current Applications in the Public Sector .................................................... 31 D. Advantages of Machine Learning in Governmental Decision-Making ....... 34 1. Accuracy. .................................................................................................. 34 2. Capacity. ................................................................................................... 35 3. Speed. ........................................................................................................ 36 4. Consistency. .............................................................................................. 37 E. Concerns About the Use of Machine Learning in Government ................... 37 1. Adequate Expertise. .................................................................................. 38 2. Data Storage and Processing Capacity .................................................... 41 3. Cybersecurity. ........................................................................................... 43 4. Privacy. ..................................................................................................... 44 5. Transparency and Explainability. ............................................................. 45 6. Bias. .......................................................................................................... 47 7. Abuse of Power. ........................................................................................ 49 III. LEGAL ISSUES WITH GOVERNMENTAL USE OF MACHINE LEARNING ............... 50 A. Delegation and Accountability .................................................................... 51 B. Procedural Due Process and Reason-Giving ................................................ 52 C. Transparency ................................................................................................ 54 D. Privacy ......................................................................................................... 55 E. Equal Protection ........................................................................................... 59 F. Overall Legal Assessment ............................................................................ 61 IV. DECIDING WHEN TO DEPLOY MACHINE LEARNING ......................................... 62 A. Multi-Factor Analysis of When to Use Machine Learning ......................... 63 B. Key Factors for Deciding Whether to Use Machine Learning .................... 66 1. Preconditions for Use. .............................................................................. 66 2. Improved Outcomes. ................................................................................. 70 3. Legal and Public Acceptance Risks. ......................................................... 72 CONCLUSION .......................................................................................................... 74 2 A Framework for Governmental Use of Machine Learning Cary Coglianese Computerized algorithms increasingly make decisions that previously had been made by humans. These new types of algorithms—known as machine learning algorithms1—have recently found themselves in use in so many products and settings that they appear to portend the reshaping of many important aspects of human life.2 With their distinctive ability to find complex patterns in large datasets, machine learning algorithms are assisting human decision-makers—or independently making forecasts and decisions—on who to hire3 or lend money,4 how to trade stocks,5 and 1 Machine learning algorithms can learn to identify patterns across the vast quantities of data that can now be stored and processed digitally, and they can do so autonomously—that is, without human specification of the form of a particular model or key variables, and subject mainly to overarching criteria or parameters to be optimized. As such, these algorithms are often discussed under the banner of “Big Data” or “artificial intelligence.” For a discussion of machine learning and how it works, see Cary Coglianese & David Lehr, Regulating by Robot: Administrative Decision Making in the Machine-Learning Era, 105 GEO. L.J. 1147, 1156-60 (2017) (hereinafter Regulating by Robot); David Lehr & Paul Ohm, Playing with the Data: What Legal Scholars Should Learn About Machine Learning, 51 UC DAVIS L. REV. 653, 669-702 (2017). Machine learning algorithms come in many forms and are referred to by a variety of terms. See Cary Coglianese & David Lehr, Transparency and Algorithmic Governance, 71 ADMIN. L. REV. 1, 2 n.2 (2019). (hereinafter Transparency) (“By ‘artificial intelligence’ and ‘machine learning,’ we refer … to a broad approach to predictive analytics captured under various umbrella terms, including ‘big data analytics,’ ‘deep learning,’
Load more

Recommended publications
  • The Availability Heuristic
    CHAPTER 11 THE AVAILABILITY HEURISTIC According to Amos Tversky and Daniel Kahneman (1974, p. 1127), the availability heuristic is a rule of thumb in which decision makers "assess the frequency of a class or the probability of an event by the ease with which instances or occurrences can be brought to mind." Usually this heuristic works quite well; all things being equal, common events are easier to remember or imagine than are uncommon events. By rely­ ing on availability to estimate frequency and probability, decision mak­ ers are able to simplify what might otherwise be very difficult judg­ ments. As with any heuristic, however, there are cases in which the general rule of thumb breaks down and leads to systematic biases. Some events are more available than others not because they tend to occur frequently or with high probability, but because they are inherently easier to think about, because they have taken place recently, because they are highly emotional, and so forth. This chapter examines three general questions: (1) What are instances in which the availability heuristic leads to biased judgments? (2) Do decision makers perceive an event as more likely after they have imagined it happening? (3) How is vivid information dif­ ferent from other information? AVAILABILITY GOES AWRY Which is a more likely cause of death in the United States-being killed by falling airplane parts or by a shark? Most people rate shark attacks as more probable than death from falling airplane parts (see Item #7 of the Reader Survey for your answer). Shark attacks certainly receive more publicity than do deaths from falling airplane parts, and they are far easier to imagine (thanks in part to movies such as Jaws).
    [Show full text]
  • Nudge Theory
    Nudge Theory http://www.businessballs.com/nudge-theory.htm Nudge theory is a flexible and modern concept for: • understanding of how people think, make decisions, and behave, • helping people improve their thinking and decisions, • managing change of all sorts, and • identifying and modifying existing unhelpful influences on people. Nudge theory was named and popularized by the 2008 book, 'Nudge: Improving Decisions About Health, Wealth, and Happiness', written by American academics Richard H Thaler and Cass R Sunstein. The book is based strongly on the Nobel prize- winning work of the Israeli-American psychologists Daniel Kahneman and Amos Tversky. This article: • reviews and explains Thaler and Sunstein's 'Nudge' concept, especially 'heuristics' (tendencies for humans to think and decide instinctively and often mistakenly) • relates 'Nudge' methods to other theories and models, and to Kahneman and Tversky's work • defines and describes additional methods of 'nudging' people and groups • extends the appreciation and application of 'Nudge' methodology to broader change- management, motivation, leadership, coaching, counselling, parenting, etc • offers 'Nudge' methods and related concepts as a 'Nudge' theory 'toolkit' so that the concept can be taught and applied in a wide range of situations involving relationships with people, and enabling people to improve their thinking and decision- making • and offers a glossary of Nudge theory and related terms 'Nudge' theory was proposed originally in US 'behavioral economics', but it can be adapted and applied much more widely for enabling and encouraging change in people, groups, or yourself. Nudge theory can also be used to explore, understand, and explain existing influences on how people behave, especially influences which are unhelpful, with a view to removing or altering them.
    [Show full text]
  • CHALK TALK Thinking About Thinking: Medical Decision Making Under the Microscope Christiana Iyasere, MD, and Douglas Wright, MD, Phd
    SGIM FORUM 2011; 34(11) CHALK TALK Thinking about Thinking: Medical Decision Making Under the Microscope Christiana Iyasere, MD, and Douglas Wright, MD, PhD Drs. Iyasere and Wright are faculty in the Inpatient Clinician Educator Service of the Department of Medicine at Massachusetts General Hospital in Boson, MA. ase: A 36-year-old African-Ameri- athletic—he graduated from college than 10 seconds and says “20,160” C can woman, healthy except for with a degree in physics and has before moving breezily along with treated hypothyroidism, visits you in completed several triathlons. Neil is a her coffee. Having Katrina’s input, are clinic complaining of six months of fa- veteran of the US Navy, where he you tempted to change your answer tigue and progressive shortness of served as fleet naval aviator and land- to questions 3a and 3b? Go ahead, breath with exertion. You thoroughly ing signal officer. Is Neil more likely to admit it. Aren’t you now more confi- interview and examine the patient. be: a) a librarian or b) an astronaut? dent that the correct answer is that Physical examination reveals conjunc- Question 2: Jot down a list of the product is closest to 20,000? tival pallor and dullness to percussion English words that begin with the let- Question 5: You have known one third of the way up both lung ter “r” (e.g. rooster). Next, jot down your medical school roommate Jus- fields. Something tells you to ask her a list of words that have an r in the tice for four years.
    [Show full text]
  • Mitigating Cognitive Bias and Design Distortion
    RSD2 Relating Systems Thinking and Design 2013 Working Paper. www.systemic-design.net Interliminal Design: Mitigating Cognitive Bias and Design Distortion 1 2 3 Andrew McCollough, PhD , DeAunne Denmark, MD, PhD , and Don Harker, MS “Design is an inquiry for action” - Harold Nelson, PhD Liminal: from the Latin word līmen, a threshold “Our values and biases are expressions of who we are. It isn’t so much that our prior commitments disable our ability to reason; it is that we need to appreciate that reasoning takes place, for the most part, in settings in which we are not neutral bystanders.” - Alva Noe In this globally interconnected and information-rich society, complexity and uncertainty are escalating at an unprecedented pace. Our most pressing problems are now considered wicked, or even super wicked, in that they commonly have incomplete, continuously changing, intricately interdependent yet contradictory requirements, urgent or emergency status, no central authority, and are often caused by the same entities charged with solving them (Rittel & Webber, 1973). Perhaps similar to philosopher Karl Jaspers’ Axial Age where ‘man asked radical questions’ and the ‘unquestioned grasp on life was loosened,’ communities of all kinds now face crises of “in-between”; necessary and large-scale dismantling of previous structures, institutions, and world-views has begun, but has yet to be replaced. Such intense periods of destruction and reconstruction, or liminality, eliminate boundary lines, reverse or dissolve social hierarchies, disrupt cultural continuity, and cast penetrating onto future outcomes. Originating in human ritual, liminal stages occur mid-transition, where initiates literally "stand at the threshold" between outmoded constructs of identity or community, and entirely new ways of being.
    [Show full text]
  • Cognitive Biases Potentially Affecting Judgment of Global Risks Forthcoming in Global Catastrophic Risks, Eds
    Cognitive biases potentially affecting judgment of global risks Forthcoming in Global Catastrophic Risks, eds. Nick Bostrom and Milan Cirkovic Draft of August 31, 2006. Eliezer Yudkowsky ([email protected]) Singularity Institute for Artificial Intelligence Palo Alto, CA Introduction1 All else being equal, not many people would prefer to destroy the world. Even faceless corporations, meddling governments, reckless scientists, and other agents of doom, require a world in which to achieve their goals of profit, order, tenure, or other villainies. If our extinction proceeds slowly enough to allow a moment of horrified realization, the doers of the deed will likely be quite taken aback on realizing that they have actually destroyed the world. Therefore I suggest that if the Earth is destroyed, it will probably be by mistake. The systematic experimental study of reproducible errors of human reasoning, and what these errors reveal about underlying mental processes, is known as the heuristics and biases program in cognitive psychology. This program has made discoveries highly relevant to assessors of global catastrophic risks. Suppose you're worried about the risk of Substance P, an explosive of planet-wrecking potency which will detonate if exposed to a strong radio signal. Luckily there's a famous expert who discovered Substance P, spent the last thirty years working with it, and knows it better than anyone else in the world. You call up the expert and ask how strong the radio signal has to be. The expert replies that the critical threshold is probably around 4,000 terawatts. "Probably?" you query. "Can you give me a 98% confidence interval?" "Sure," replies the expert.
    [Show full text]
  • Heuristics & Biases Simplified
    University of Pennsylvania ScholarlyCommons Master of Behavioral and Decision Sciences Capstones Behavioral and Decision Sciences Program 8-9-2019 Heuristics & Biases Simplified: Easy ot Understand One-Pagers that Use Plain Language & Examples to Simplify Human Judgement and Decision-Making Concepts Brittany Gullone University of Pennsylvania Follow this and additional works at: https://repository.upenn.edu/mbds Part of the Social and Behavioral Sciences Commons Gullone, Brittany, "Heuristics & Biases Simplified: Easy to Understand One-Pagers that Use Plain Language & Examples to Simplify Human Judgement and Decision-Making Concepts" (2019). Master of Behavioral and Decision Sciences Capstones. 7. https://repository.upenn.edu/mbds/7 The final capstone research project for the Master of Behavioral and Decision Sciences is an independent study experience. The paper is expected to: • Present a position that is unique, original and directly applies to the student's experience; • Use primary sources or apply to a primary organization/agency; • Conform to the style and format of excellent academic writing; • Analyze empirical research data that is collected by the student or that has already been collected; and • Allow the student to demonstrate the competencies gained in the master’s program. This paper is posted at ScholarlyCommons. https://repository.upenn.edu/mbds/7 For more information, please contact [email protected]. Heuristics & Biases Simplified: Easy ot Understand One-Pagers that Use Plain Language & Examples to Simplify Human Judgement and Decision-Making Concepts Abstract Behavioral Science is a new and quickly growing field of study that has found ways of capturing readers’ attention across a variety of industries. The popularity of this field has led to a wealth of terms, concepts, and materials that describe human behavior and decision making.
    [Show full text]
  • A Disinformation-Misinformation Ecology: the Case of Trump Thomas J
    Chapter A Disinformation-Misinformation Ecology: The Case of Trump Thomas J. Froehlich Abstract This paper lays out many of the factors that make disinformation or misinformation campaigns of Trump successful. By all rational standards, he is unfit for office, a compulsive liar, incompetent, arrogant, ignorant, mean, petty, and narcissistic. Yet his approval rating tends to remain at 40%. Why do rational assessments of his presidency fail to have any traction? This paper looks at the con- flation of knowledge and beliefs in partisan minds, how beliefs lead to self-decep- tion and social self-deception and how they reinforce one another. It then looks at psychological factors, conscious and unconscious, that predispose partisans to pursue partisan sources of information and reject non-partisan sources. It then explains how these factors sustain the variety and motivations of Trump supporters’ commitment to Trump. The role of cognitive authorities like Fox News and right-wing social media sites are examined to show how the power of these media sources escalates and reinforces partisan views and the rejection of other cognitive authorities. These cognitive authorities also use emotional triggers to inflame Trump supporters, keeping them addicted by feeding their anger, resentment, or self-righteousness. The paper concludes by discussing the dynamics of the Trump disinformation- misinformation ecology, creating an Age of Inflamed Grievances. Keywords: Trumpism, disinformation, cognitive authority, Fox News, social media, propaganda, inflamed grievances, psychology of disinformation, Donald Trump, media, self-deception, social self-deception 1. Introduction This paper investigates how disinformation-misinformation campaigns, particularly in the political arena, succeed and why they are so hard to challenge, defeat, or deflect.
    [Show full text]
  • Wishful Thinking About the Future: Does Desire Impact Optimism? Zlatan Krizan1* and Paul D
    Social and Personality Psychology Compass 3/3 (2009): 227–243, 10.1111/j.1751-9004.2009.00169.x WishfulBlackwellOxford,SPCOSocial1751-9004©Journal16910.1111/j.1751-9004.2009.00169.xMarch0227???243???Original 2009 and2009 Thinking UKTheCompilationArticle Publishing Personality Authors ©Ltd Psychology2009 Blackwell Compass Publishing Ltd. Wishful Thinking about the Future: Does Desire Impact Optimism? Zlatan Krizan1* and Paul D. Windschitl2 1 Iowa State University 2 University of Iowa Abstract The notion that desire for an outcome inflates optimism about that outcome has been dubbed the desirability bias or wishful thinking. In this paper, we discuss the importance of distinguishing wishful thinking from the more general concept of motivated reasoning, and we explain why documenting overoptimism or correla- tions between preferences and optimism is not sufficient to infer a desirability bias. Then, we discuss results from a review and meta-analysis of the experimental litera- ture on wishful thinking. These findings, in conjunction with more recent work, not only highlight important moderators and mediators of the desirability bias but also point out limitations of the empirical research on the bias. These results also reveal an important difference between how likelihood judgments and discrete outcome predictions respond to desirability of outcomes. We conclude by presenting avenues for future research useful for understanding wishful thinking’s manifesta- tion in everyday environments and its integration with related phenomena. —A soon-to-be bride and groom who hope the weather will cooperate on their wedding day. —An Illinois grade-schooler who thinks it would be ‘pretty cool’ if Chicago was indeed selected as the host city for the 2016 Olympics.
    [Show full text]
  • The Future of Global Affairs Managing Discontinuity, Disruption and Destruction
    The Future of Global Affairs Managing Discontinuity, Disruption and Destruction Edited by Christopher Ankersen · Waheguru Pal Singh Sidhu Foreword by Helen Clark and Vera Jelinek DOWNLOAD CSS Notes, Books, MCQs, Magazines www.thecsspoint.com Download CSS Notes Download CSS Books Download CSS Magazines Download CSS MCQs Download CSS Past Papers The CSS Point, Pakistan’s The Best Online FREE Web source for All CSS Aspirants. Email: [email protected] BUY CSS / PMS / NTS & GENERAL KNOWLEDGE BOOKS ONLINE CASH ON DELIVERY ALL OVER PAKISTAN Visit Now: WWW.CSSBOOKS.NET For Oder & Inquiry Call/SMS/WhatsApp 0333 6042057 – 0726 540141 Join CSS Point Forum Today Discuss, Learn and Explore The World of CSS www.csspoint.com.pk For any Query, Please Call/WhatsApp 03336042057 - 0726540141 ENGLISH PRECIS & COMPOSITION HAFIZ KARIM DAD CHUGTAI For Order Call/WhatsApp 03336042057 - 0726540141 The Future of Global Affairs “The Future of Global Affairs is an exciting and balanced contribution to the debate about the potential trajectory of a world in flux. Tinged more with concern than utopian optimism, this volume captures a panoramic view of our cacophonous and disorderly world that is on the verge of disequilibrium and potential destruction unless key political actors, institutions, and processes can find a way to adapt global affairs to an increasingly plurilateral and intermestic era. A must-read for serious IR scholars; written in a way that is understandable to the lay person.” —W. Andy Knight, Ph.D., FRSC, University of Alberta “What is the future of the state? Is multilateralism overrated as an answer to the diverse ills plaguing the global system? How might climate change disrupt geopolitical alliances that have been essential to peace in different regions of the world? Inspired by these and other urgent questions, this simultaneously illuminating and profoundly unsettling book traces the contours of the ascendant geopolitics, illustrating how one class of events can have contrasting resonances and implications around the world.
    [Show full text]
  • Cognitive Biases in Fitness to Practise Decision-Making
    ADVICE ON BIASES IN FITNESS TO PRACTISE DECISION-MAKING IN ACCEPTED OUTCOME VERSUS PANEL MODELS FOR THE PROFESSIONAL STANDARDS AUTHORITY Prepared by: LESLIE CUTHBERT Date: 31 March 2021 Contents I. Introduction ................................................................................................. 2 Background .................................................................................................. 2 Purpose of the advice.................................................................................... 3 Summary of conclusions ................................................................................ 4 Documents considered .................................................................................. 4 Technical terms and explanations ................................................................... 5 Disclaimer .................................................................................................... 5 II. How biases might affect the quality of decision-making in the AO model, as compared to the panel model .............................................................................. 6 Understanding cognitive biases in fitness to practise decision-making ............... 6 Individual Decision Maker (e.g. a Case Examiner) in AO model ......................... 8 Multiple Case Examiners .............................................................................. 17 Group Decision Maker in a Panel model of 3 or more members ....................... 22 III. Assessment of the impact of these biases,
    [Show full text]
  • Psychology 36. Thinking, Judgment, and Decision Making Barry Schwartz Fall, 2006
    Psychology 36. Thinking, Judgment, and Decision Making Barry Schwartz Fall, 2006 In the modern world, people are faced with difficult decisions on an almost daily basis. From consumer purchases to college choices to romantic commitments to health care options to career paths, decisions must be made in the face of plentiful but incomplete information and uncertain outcomes. In this course, we will examine both how people should go about making decisions and how people do go about making decisions. We will see that there is often quite a substantial gap between what people should do and what people do do, and we will examine what, if anything, can be done to close that gap. The aim of the course is both to familiarize you with exciting contemporary research on judgment and decision making, and to give you tools that might help you to improve your own decision making. What makes the study of judgment and decision making so interesting and exciting to me is that it studies basic psychological processes in a way that is extremely closely connected to real life application—both to personal life and to public policy. By studying how people formulate goals, determine values, gather and evaluate evidence, think about uncertainty, make choices, think about the results of those choices, and revise their goals and values, we will simultaneously (I hope) make ourselves more conscious of our own decision making processes so that we can improve upon them. Requirements The course will include a midterm exam, a final exam, and a term paper (8 pages). The term paper assignment will be described in class.
    [Show full text]
  • Heuristics and Public Policy: Decision Making Under Bounded Rationality
    ISSN 1936-5349 (print) ISSN 1936-5357 (online) HARVARD JOHN M. OLIN CENTER FOR LAW, ECONOMICS, AND BUSINESS HEURISTICS AND PUBLIC POLICY: DECISION MAKING UNDER BOUNDED RATIONALITY Sanjit Dhami Ali al-Nowaihi Cass R. Sunstein Discussion Paper No. 963 06/2018 Harvard Law School Cambridge, MA 02138 This paper can be downloaded without charge from: The Harvard John M. Olin Discussion Paper Series: http://www.law.harvard.edu/programs/olin_center The Social Science Research Network Electronic Paper Collection: https://ssrn.com/abstract=3198739 Heuristics and Public Policy: Decision Making Under Bounded Rationality Sanjit Dhami Ali al-NowaihiyCass R. Sunsteinz 18 June 2018 Abstract How do human beings make decisions when, as the evidence indicates, the as- sumptions of the Bayesian rationality approach in economics do not hold? Do human beings optimize, or can they? Several decades of research have shown that people possess a toolkit of heuristics to make decisions under certainty, risk, subjective uncertainty, and true uncertainty (or Knightian uncertainty). We outline recent advances in knowledge about the use of heuristics and departures from Bayesian rationality, with particular emphasis on growing formalization of those departures, which add necessary precision. We also explore the relationship between bounded rationality and libertarian paternalism, or nudges, and show that some recent ob- jections, founded on psychological work on the usefulness of certain heuristics, are based on serious misunderstandings. Keywords: Heuristics; biases; Bayesian rationality; ecological rationality; true uncer- tainty; libertarian paternalism. Professor of Economics, University of Leicester, United Kingdom. yProfessor Emeritus of Economics, University of Leicester, United Kingdom. zRobert Walmsley University Professor, Harvard University, USA.
    [Show full text]